Home

Context Navigation

source: palm/trunk/SOURCE/init_pegrid.f90 @ 4554

Last change on this file since 4554 was 4461, checked in by raasch, 5 years ago
extensions to allow usage of alternative communicators in exchange_horiz
Property svn:keywords set to `Id`
File size: 53.6 KB

Line
1	!> @file init_pegrid.f90
2	!------------------------------------------------------------------------------!
3	! This file is part of the PALM model system.
4	!
5	! PALM is free software: you can redistribute it and/or modify it under the
6	! terms of the GNU General Public License as published by the Free Software
7	! Foundation, either version 3 of the License, or (at your option) any later
8	! version.
9	!
10	! PALM is distributed in the hope that it will be useful, but WITHOUT ANY
11	! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
12	! A PARTICULAR PURPOSE. See the GNU General Public License for more details.
13	!
14	! You should have received a copy of the GNU General Public License along with
15	! PALM. If not, see <http://www.gnu.org/licenses/>.
16	!
17	! Copyright 1997-2020 Leibniz Universitaet Hannover
18	!------------------------------------------------------------------------------!
19	!
20	! Current revisions:
21	! ------------------
22	!
23	!
24	! Former revisions:
25	! -----------------
26	! $Id: init_pegrid.f90 4461 2020-03-12 16:51:59Z eckhard $
27	! communicator configurations for four virtual pe grids defined
28	!
29	! 4444 2020-03-05 15:59:50Z raasch
30	! bugfix: cpp-directives for serial mode added
31	!
32	! 4360 2020-01-07 11:25:50Z suehring
33	! changed message PA0467
34	!
35	! 4264 2019-10-15 16:00:23Z scharf
36	! corrected error message string
37	!
38	! 4241 2019-09-27 06:32:47Z raasch
39	! Check added to ensure that subdomain grid has at least the size as given by the number
40	! of ghost points
41	!
42	! 4182 2019-08-22 15:20:23Z scharf
43	! Corrected "Former revisions" section
44	!
45	! 4045 2019-06-21 10:58:47Z raasch
46	! bugfix: kind attribute added to nint function to allow for large integers which may appear in
47	! case of default recycling width and small grid spacings
48	!
49	! 3999 2019-05-23 16:09:37Z suehring
50	! Spend 3 ghost points also in case of pw-scheme when nesting is applied
51	!
52	! 3897 2019-04-15 11:51:14Z suehring
53	! Minor revision of multigrid check; give warning instead of an abort.
54	!
55	! 3890 2019-04-12 15:59:20Z suehring
56	! Check if grid coarsening is possible on subdomain, in order to avoid that
57	! multigrid approach effectively reduces to a Gauss-Seidel scheme.
58	!
59	! 3885 2019-04-11 11:29:34Z kanani
60	! Changes related to global restructuring of location messages and introduction
61	! of additional debug messages
62	!
63	! 3884 2019-04-10 13:31:55Z Giersch
64	! id_recycling is only calculated in case of tubulent inflow
65	!
66	! 3761 2019-02-25 15:31:42Z raasch
67	! unused variable removed
68	!
69	! 3655 2019-01-07 16:51:22Z knoop
70	! variables documented
71	!
72	! Revision 1.1 1997/07/24 11:15:09 raasch
73	! Initial revision
74	!
75	!
76	! Description:
77	! ------------
78	!> Determination of the virtual processor topology (if not prescribed by the
79	!> user)and computation of the grid point number and array bounds of the local
80	!> domains.
81	!> @todo: remove MPI-data types for 2D exchange on coarse multigrid level (not
82	!> used any more)
83	!------------------------------------------------------------------------------!
84	SUBROUTINE init_pegrid
85
86
87	USE control_parameters, &
88	ONLY: bc_dirichlet_l, bc_dirichlet_n, bc_dirichlet_r, bc_dirichlet_s, &
89	bc_lr, bc_ns, bc_radiation_l, bc_radiation_n, bc_radiation_r, &
90	bc_radiation_s, &
91	grid_level, grid_level_count, maximum_grid_level, &
92	message_string, mg_switch_to_pe0_level, &
93	psolver
94
95
96	#if defined( __parallel )
97	USE control_parameters, &
98	ONLY: coupling_mode, coupling_topology, gathered_size, momentum_advec, &
99	outflow_source_plane, recycling_width, scalar_advec, subdomain_size, &
100	turbulent_inflow, turbulent_outflow, y_shift
101
102	USE grid_variables, &
103	ONLY: dx
104	#endif
105
106	USE indices, &
107	ONLY: nnx, nny, nnz, nx, nxl, nxl_mg, &
108	nxlu, nxr, nxr_mg, ny, nyn, nyn_mg, nys, nys_mg, &
109	nysv, nz, nzb, nzt, nzt_mg, wall_flags_1, wall_flags_2, &
110	wall_flags_3, wall_flags_4, wall_flags_5, wall_flags_6, &
111	wall_flags_7, wall_flags_8, wall_flags_9, wall_flags_10
112
113	#if defined( __parallel )
114	USE indices, &
115	ONLY: mg_loc_ind, nbgp, nx_a, nx_o, ny_a, ny_o
116	#endif
117
118	USE kinds
119
120	USE pegrid
121
122	#if defined( __parallel )
123	USE pmc_interface, &
124	ONLY: nested_run
125
126	USE spectra_mod, &
127	ONLY: calculate_spectra
128
129	USE synthetic_turbulence_generator_mod, &
130	ONLY: id_stg_left, id_stg_north, id_stg_right, id_stg_south, &
131	use_syn_turb_gen
132	#endif
133
134	USE transpose_indices, &
135	ONLY: nxl_y, nxl_z, nxr_y, nxr_z, nyn_x, nyn_z, nys_x,&
136	nys_z, nzb_x, nzb_y, nzt_x, nzt_y
137
138	#if defined( __parallel )
139	USE transpose_indices, &
140	ONLY: nxl_yd, nxr_yd, nzb_yd, nzt_yd
141
142	USE vertical_nesting_mod, &
143	ONLY: vnested, vnest_init_pegrid_domain, vnest_init_pegrid_rank
144	#endif
145
146	IMPLICIT NONE
147
148	INTEGER(iwp) :: i !< running index over number of processors or number of multigrid level
149	#if defined( __parallel )
150	INTEGER(iwp) :: id_inflow_l !< ID indicating processors located at the left inflow boundary
151	INTEGER(iwp) :: id_outflow_l !< local value of id_outflow
152	INTEGER(iwp) :: id_outflow_source_l !< local value of id_outflow_source
153	INTEGER(iwp) :: id_recycling_l !< ID indicating processors located at the recycling plane
154	INTEGER(iwp) :: id_stg_left_l !< left lateral boundary local core id in case of turbulence generator
155	INTEGER(iwp) :: id_stg_north_l !< north lateral boundary local core id in case of turbulence generator
156	INTEGER(iwp) :: id_stg_right_l !< right lateral boundary local core id in case of turbulence generator
157	INTEGER(iwp) :: id_stg_south_l !< south lateral boundary local core id in case of turbulence generator
158	INTEGER(iwp) :: ind(5) !< array containing the subdomain bounds
159	#endif
160	INTEGER(iwp) :: j !< running index, used for various loops
161	INTEGER(iwp) :: k !< number of vertical grid points in different multigrid level
162	INTEGER(iwp) :: maximum_grid_level_l !< maximum number of grid level without switching to PE 0
163	INTEGER(iwp) :: mg_levels_x !< maximum number of grid level allowed along x-direction
164	INTEGER(iwp) :: mg_levels_y !< maximum number of grid level allowed along y-direction
165	INTEGER(iwp) :: mg_levels_z !< maximum number of grid level allowed along z-direction
166	INTEGER(iwp) :: mg_switch_to_pe0_level_l !< maximum number of grid level with switching to PE 0
167	#if defined( __parallel )
168	INTEGER(iwp) :: nnx_y !< quotient of number of grid points along x-direction and number of PEs used along y-direction
169	INTEGER(iwp) :: nny_x !< quotient of number of grid points along y-direction and number of PEs used along x-direction
170	INTEGER(iwp) :: nny_z !< quotient of number of grid points along y-direction and number of PEs used along x-direction
171	INTEGER(iwp) :: nnz_x !< quotient of number of grid points along z-direction and number of PEs used along x-direction
172	INTEGER(iwp) :: nnz_y !< quotient of number of grid points along z-direction and number of PEs used along x-direction
173	INTEGER(iwp) :: numproc_sqr !< square root of the number of processors
174	#endif
175	INTEGER(iwp) :: nxl_l !< lower index bound along x-direction on subdomain and different multigrid level
176	INTEGER(iwp) :: nxr_l !< upper index bound along x-direction on subdomain and different multigrid level
177	INTEGER(iwp) :: nyn_l !< lower index bound along y-direction on subdomain and different multigrid level
178	INTEGER(iwp) :: nys_l !< upper index bound along y-direction on subdomain and different multigrid level
179	#if defined( __parallel )
180	INTEGER(iwp) :: nzb_l !< lower index bound along z-direction on subdomain and different multigrid level
181	#endif
182	INTEGER(iwp) :: nzt_l !< upper index bound along z-direction on subdomain and different multigrid level
183	!$ INTEGER(iwp) :: omp_get_num_threads !< number of OpenMP threads
184
185	#if defined( __parallel )
186	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: ind_all !< dummy array containing index bounds on subdomain, used for gathering
187	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxlf !< lower index bound allong x-direction for every PE
188	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nxrf !< upper index bound allong x-direction for every PE
189	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nynf !< lower index bound allong y-direction for every PE
190	INTEGER(iwp), DIMENSION(:), ALLOCATABLE :: nysf !< lower index bound allong y-direction for every PE
191
192	INTEGER(iwp), DIMENSION(2) :: pdims_remote !< number of PEs used for coupled model (only in atmospher-ocean coupling)
193	INTEGER(iwp) :: lcoord(2) !< PE coordinates of left neighbor along x and y
194	INTEGER(iwp) :: rcoord(2) !< PE coordinates of right neighbor along x and y
195	#endif
196
197	!
198	!-- Get the number of OpenMP threads
199	!$OMP PARALLEL
200	!$ threads_per_task = omp_get_num_threads()
201	!$OMP END PARALLEL
202
203
204	#if defined( __parallel )
205
206	CALL location_message( 'creating virtual PE grids + MPI derived data types', 'start' )
207
208	!
209	!-- Determine the processor topology or check it, if prescribed by the user
210	IF ( npex == -1 .AND. npey == -1 ) THEN
211
212	!
213	!-- Automatic determination of the topology
214	numproc_sqr = SQRT( REAL( numprocs, KIND=wp ) )
215	pdims(1) = MAX( numproc_sqr , 1 )
216	DO WHILE ( MOD( numprocs , pdims(1) ) /= 0 )
217	pdims(1) = pdims(1) - 1
218	ENDDO
219	pdims(2) = numprocs / pdims(1)
220
221	ELSEIF ( npex /= -1 .AND. npey /= -1 ) THEN
222
223	!
224	!-- Prescribed by user. Number of processors on the prescribed topology
225	!-- must be equal to the number of PEs available to the job
226	IF ( ( npex * npey ) /= numprocs ) THEN
227	WRITE( message_string, * ) 'number of PEs of the prescribed ', &
228	'topology (', npex*npey,') does not match & the number of ', &
229	'PEs available to the job (', numprocs, ')'
230	CALL message( 'init_pegrid', 'PA0221', 1, 2, 0, 6, 0 )
231	ENDIF
232	pdims(1) = npex
233	pdims(2) = npey
234
235	ELSE
236	!
237	!-- If the processor topology is prescribed by the user, the number of
238	!-- PEs must be given in both directions
239	message_string = 'if the processor topology is prescribed by th' // &
240	'e user & both values of "npex" and "npey" must be given' // &
241	' in the &NAMELIST-parameter file'
242	CALL message( 'init_pegrid', 'PA0222', 1, 2, 0, 6, 0 )
243
244	ENDIF
245
246	!
247	!-- Create four default MPI communicators for the 2d virtual PE grid. One of them will be used
248	!-- as the main communicator for this run, while others might be used for specific quantities like
249	!-- aerosol, chemical species, or passive scalars), if their horizontal boundary conditions shall
250	!-- be different from those of the other quantities (e.g. non-cyclic conditions for aerosols, and
251	!-- cyclic conditions for all others).
252	DO i = 1, 4
253
254	IF ( i == 1 ) cyclic = (/ .TRUE., .TRUE. /) ! cyclic along x and y
255	IF ( i == 2 ) cyclic = (/ .TRUE., .FALSE. /) ! cyclic along x
256	IF ( i == 3 ) cyclic = (/ .FALSE., .TRUE. /) ! cyllic along y
257	IF ( i == 4 ) cyclic = (/ .FALSE., .FALSE. /) ! non-cyclic
258
259	CALL MPI_CART_CREATE( comm_palm, ndim, pdims, cyclic, reorder, &
260	communicator_configurations(i)%mpi_communicator, ierr )
261
262	CALL MPI_CART_SHIFT( communicator_configurations(i)%mpi_communicator, 0, 1, &
263	communicator_configurations(i)%pleft, &
264	communicator_configurations(i)%pright, ierr )
265
266	CALL MPI_CART_SHIFT( communicator_configurations(i)%mpi_communicator, 1, 1, &
267	communicator_configurations(i)%psouth, &
268	communicator_configurations(i)%pnorth, ierr )
269
270	ENDDO
271
272	!
273	!-- If necessary, set horizontal boundary conditions to non-cyclic
274	IF ( bc_lr /= 'cyclic' ) cyclic(1) = .FALSE.
275	IF ( bc_ns /= 'cyclic' ) cyclic(2) = .FALSE.
276
277
278	!
279	!-- Set the main communicator (virtual pe grid) for this run
280	IF ( bc_lr == 'cyclic' .AND. bc_ns == 'cyclic' ) i = 1
281	IF ( bc_lr == 'cyclic' .AND. bc_ns /= 'cyclic' ) i = 2
282	IF ( bc_lr /= 'cyclic' .AND. bc_ns == 'cyclic' ) i = 3
283	IF ( bc_lr /= 'cyclic' .AND. bc_ns /= 'cyclic' ) i = 4
284
285	comm2d = communicator_configurations(i)%mpi_communicator
286	pleft = communicator_configurations(i)%pleft
287	pright = communicator_configurations(i)%pright
288	psouth = communicator_configurations(i)%psouth
289	pnorth = communicator_configurations(i)%pnorth
290
291	!
292	!-- Set rank and coordinates of the main communicator
293	CALL MPI_COMM_RANK( comm2d, myid, ierr )
294	WRITE (myid_char,'(''_'',I6.6)') myid
295
296	CALL MPI_CART_COORDS( comm2d, myid, ndim, pcoord, ierr )
297
298	!
299	!-- In case of cyclic boundary conditions, a y-shift at the boundaries in
300	!-- x-direction can be introduced via parameter y_shift. The shift is done
301	!-- by modifying the processor grid in such a way that processors located
302	!-- at the x-boundary communicate across it to processors with y-coordinate
303	!-- shifted by y_shift relative to their own. This feature can not be used
304	!-- in combination with an fft pressure solver. It has been implemented to
305	!-- counter the effect of streak structures in case of cyclic boundary
306	!-- conditions. For a description of these see Munters
307	!-- (2016; dx.doi.org/10.1063/1.4941912)
308	!--
309	!-- Get coordinates of left and right neighbor on PE grid
310	IF ( y_shift /= 0 ) THEN
311	IF ( bc_lr == 'cyclic' ) THEN
312	IF ( TRIM( psolver ) /= 'multigrid' .AND. &
313	TRIM( psolver ) /= 'multigrid_noopt') &
314	THEN
315	message_string = 'y_shift /= 0 requires a multigrid pressure solver '
316	CALL message( 'check_parameters', 'PA0468', 1, 2, 0, 6, 0 )
317	ENDIF
318
319	CALL MPI_CART_COORDS( comm2d, pright, ndim, rcoord, ierr )
320	CALL MPI_CART_COORDS( comm2d, pleft, ndim, lcoord, ierr )
321
322	!
323	!-- If the x(y)-coordinate of the right (left) neighbor is smaller (greater)
324	!-- than that of the calling process, then the calling process is located on
325	!-- the right (left) boundary of the processor grid. In that case,
326	!-- the y-coordinate of that neighbor is increased (decreased) by y_shift.
327	!-- The rank of the process with that coordinate is then inquired and the
328	!-- neighbor rank for MPI_SENDRECV, pright (pleft) is set to it.
329	!-- In this way, the calling process receives a new right (left) neighbor
330	!-- for all future MPI_SENDRECV calls. That neighbor has a y-coordinate
331	!-- of y+(-)y_shift, where y is the original right (left) neighbor's
332	!-- y-coordinate. The modulo-operation ensures that if the neighbor's
333	!-- y-coordinate exceeds the grid-boundary, it will be relocated to
334	!-- the opposite part of the grid cyclicly.
335	IF ( rcoord(1) < pcoord(1) ) THEN
336	rcoord(2) = MODULO( rcoord(2) + y_shift, pdims(2) )
337	CALL MPI_CART_RANK( comm2d, rcoord, pright, ierr )
338	ENDIF
339
340	IF ( lcoord(1) > pcoord(1) ) THEN
341	lcoord(2) = MODULO( lcoord(2) - y_shift, pdims(2) )
342	CALL MPI_CART_RANK( comm2d, lcoord, pleft, ierr )
343	ENDIF
344
345	ELSE
346	!
347	!-- y-shift for non-cyclic boundary conditions is only implemented
348	!-- for the turbulence recycling method in inflow_turbulence.f90
349	IF ( .NOT. turbulent_inflow ) THEN
350	message_string = 'y_shift /= 0 is only allowed for cyclic ' // &
351	'boundary conditions in both directions ' // &
352	'or with turbulent_inflow == .TRUE.'
353	CALL message( 'check_parameters', 'PA0467', 1, 2, 0, 6, 0 )
354	ENDIF
355	ENDIF
356	ENDIF
357
358	!
359	!-- Vertical nesting: store four lists that identify partner ranks to exchange
360	!-- data
361	IF ( vnested ) CALL vnest_init_pegrid_rank
362
363	!
364	!-- Determine sub-topologies for transpositions
365	!-- Transposition from z to x:
366	remain_dims(1) = .TRUE.
367	remain_dims(2) = .FALSE.
368	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dx, ierr )
369	CALL MPI_COMM_RANK( comm1dx, myidx, ierr )
370	!
371	!-- Transposition from x to y
372	remain_dims(1) = .FALSE.
373	remain_dims(2) = .TRUE.
374	CALL MPI_CART_SUB( comm2d, remain_dims, comm1dy, ierr )
375	CALL MPI_COMM_RANK( comm1dy, myidy, ierr )
376
377
378	!
379	!-- Calculate array bounds along x-direction for every PE.
380	ALLOCATE( nxlf(0:pdims(1)-1), nxrf(0:pdims(1)-1), nynf(0:pdims(2)-1), &
381	nysf(0:pdims(2)-1) )
382
383	IF ( MOD( nx+1 , pdims(1) ) /= 0 ) THEN
384	WRITE( message_string, * ) 'x-direction: gridpoint number (',nx+1,') ', &
385	'is not an& integral multiple of the number',&
386	' of processors (',pdims(1),')'
387	CALL message( 'init_pegrid', 'PA0225', 1, 2, 0, 6, 0 )
388	ELSE
389	nnx = ( nx + 1 ) / pdims(1)
390	ENDIF
391
392	!
393	!-- Left and right array bounds, number of gridpoints
394	DO i = 0, pdims(1)-1
395	nxlf(i) = i * nnx
396	nxrf(i) = ( i + 1 ) * nnx - 1
397	ENDDO
398
399	!
400	!-- Calculate array bounds in y-direction for every PE.
401	IF ( MOD( ny+1 , pdims(2) ) /= 0 ) THEN
402	WRITE( message_string, * ) 'y-direction: gridpoint number (',ny+1,') ', &
403	'is not an& integral multiple of the number',&
404	' of processors (',pdims(2),')'
405	CALL message( 'init_pegrid', 'PA0227', 1, 2, 0, 6, 0 )
406	ELSE
407	nny = ( ny + 1 ) / pdims(2)
408	ENDIF
409
410	!
411	!-- South and north array bounds
412	DO j = 0, pdims(2)-1
413	nysf(j) = j * nny
414	nynf(j) = ( j + 1 ) * nny - 1
415	ENDDO
416
417	!
418	!-- Local array bounds of the respective PEs
419	nxl = nxlf(pcoord(1))
420	nxr = nxrf(pcoord(1))
421	nys = nysf(pcoord(2))
422	nyn = nynf(pcoord(2))
423	nzb = 0
424	nzt = nz
425	nnz = nz
426
427	!
428	!-- Set switches to define if the PE is situated at the border of the virtual
429	!-- processor grid
430	IF ( nxl == 0 ) left_border_pe = .TRUE.
431	IF ( nxr == nx ) right_border_pe = .TRUE.
432	IF ( nys == 0 ) south_border_pe = .TRUE.
433	IF ( nyn == ny ) north_border_pe = .TRUE.
434
435	!
436	!-- Calculate array bounds and gridpoint numbers for the transposed arrays
437	!-- (needed in the pressure solver)
438	!-- For the transposed arrays, cyclic boundaries as well as top and bottom
439	!-- boundaries are omitted, because they are obstructive to the transposition
440
441	!
442	!-- 1. transposition z --> x
443	!-- This transposition is not neccessary in case of a 1d-decomposition along x
444	IF ( psolver == 'poisfft' .OR. calculate_spectra ) THEN
445
446	IF ( pdims(2) /= 1 ) THEN
447	IF ( MOD( nz , pdims(1) ) /= 0 ) THEN
448	WRITE( message_string, * ) 'transposition z --> x:& ', &
449	'nz=',nz,' is not an integral multiple ',&
450	'of pdims(1)=',pdims(1)
451	CALL message( 'init_pegrid', 'PA0230', 1, 2, 0, 6, 0 )
452	ENDIF
453	ENDIF
454
455	nys_x = nys
456	nyn_x = nyn
457	nny_x = nny
458	nnz_x = nz / pdims(1)
459	nzb_x = 1 + myidx * nnz_x
460	nzt_x = ( myidx + 1 ) * nnz_x
461	sendrecvcount_zx = nnx * nny * nnz_x
462
463	ENDIF
464
465
466	IF ( psolver == 'poisfft' ) THEN
467	!
468	!-- 2. transposition x --> y
469	IF ( MOD( nx+1 , pdims(2) ) /= 0 ) THEN
470	WRITE( message_string, * ) 'transposition x --> y:& ', &
471	'nx+1=',nx+1,' is not an integral ', &
472	'multiple of pdims(2)=',pdims(2)
473	CALL message( 'init_pegrid', 'PA0231', 1, 2, 0, 6, 0 )
474	ENDIF
475
476	nnz_y = nnz_x
477	nzb_y = nzb_x
478	nzt_y = nzt_x
479	nnx_y = (nx+1) / pdims(2)
480	nxl_y = myidy * nnx_y
481	nxr_y = ( myidy + 1 ) * nnx_y - 1
482	sendrecvcount_xy = nnx_y * nny_x * nnz_y
483	!
484	!-- 3. transposition y --> z
485	!-- (ELSE: x --> y in case of 1D-decomposition along x)
486	nxl_z = nxl_y
487	nxr_z = nxr_y
488	nny_z = (ny+1) / pdims(1)
489	nys_z = myidx * nny_z
490	nyn_z = ( myidx + 1 ) * nny_z - 1
491	sendrecvcount_yz = nnx_y * nny_z * nnz_y
492
493	IF ( pdims(2) /= 1 ) THEN
494	!
495	!-- y --> z
496	!-- This transposition is not neccessary in case of a 1d-decomposition
497	!-- along x, except that the uptream-spline method is switched on
498	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
499	WRITE( message_string, * ) 'transposition y --> z:& ', &
500	'ny+1=',ny+1,' is not an integral ', &
501	'multiple of pdims(1)=',pdims(1)
502	CALL message( 'init_pegrid', 'PA0232', 1, 2, 0, 6, 0 )
503	ENDIF
504
505	ELSE
506	!
507	!-- x --> y
508	!-- This condition must be fulfilled for a 1D-decomposition along x
509	IF ( MOD( ny+1 , pdims(1) ) /= 0 ) THEN
510	WRITE( message_string, * ) 'transposition x --> y:& ', &
511	'ny+1=',ny+1,' is not an integral ', &
512	'multiple of pdims(1)=',pdims(1)
513	CALL message( 'init_pegrid', 'PA0233', 1, 2, 0, 6, 0 )
514	ENDIF
515
516	ENDIF
517
518	ENDIF
519
520	!
521	!-- Indices for direct transpositions z --> y (used for calculating spectra)
522	IF ( calculate_spectra ) THEN
523	IF ( MOD( nz, pdims(2) ) /= 0 ) THEN
524	WRITE( message_string, * ) 'direct transposition z --> y (needed ', &
525	'for spectra):& nz=',nz,' is not an ', &
526	'integral multiple of pdims(2)=',pdims(2)
527	CALL message( 'init_pegrid', 'PA0234', 1, 2, 0, 6, 0 )
528	ELSE
529	nxl_yd = nxl
530	nxr_yd = nxr
531	nzb_yd = 1 + myidy * ( nz / pdims(2) )
532	nzt_yd = ( myidy + 1 ) * ( nz / pdims(2) )
533	sendrecvcount_zyd = nnx * nny * ( nz / pdims(2) )
534	ENDIF
535	ENDIF
536
537	IF ( psolver == 'poisfft' .OR. calculate_spectra ) THEN
538	!
539	!-- Indices for direct transpositions y --> x
540	!-- (they are only possible in case of a 1d-decomposition along x)
541	IF ( pdims(2) == 1 ) THEN
542	nny_x = nny / pdims(1)
543	nys_x = myid * nny_x
544	nyn_x = ( myid + 1 ) * nny_x - 1
545	nzb_x = 1
546	nzt_x = nz
547	sendrecvcount_xy = nnx * nny_x * nz
548	ENDIF
549
550	ENDIF
551
552	IF ( psolver == 'poisfft' ) THEN
553	!
554	!-- Indices for direct transpositions x --> y
555	!-- (they are only possible in case of a 1d-decomposition along y)
556	IF ( pdims(1) == 1 ) THEN
557	nnx_y = nnx / pdims(2)
558	nxl_y = myid * nnx_y
559	nxr_y = ( myid + 1 ) * nnx_y - 1
560	nzb_y = 1
561	nzt_y = nz
562	sendrecvcount_xy = nnx_y * nny * nz
563	ENDIF
564
565	ENDIF
566
567	!
568	!-- Arrays for storing the array bounds are needed any more
569	DEALLOCATE( nxlf , nxrf , nynf , nysf )
570
571
572	!
573	!-- Collect index bounds from other PEs (to be written to restart file later)
574	ALLOCATE( hor_index_bounds(4,0:numprocs-1) )
575
576	IF ( myid == 0 ) THEN
577
578	hor_index_bounds(1,0) = nxl
579	hor_index_bounds(2,0) = nxr
580	hor_index_bounds(3,0) = nys
581	hor_index_bounds(4,0) = nyn
582
583	!
584	!-- Receive data from all other PEs
585	DO i = 1, numprocs-1
586	CALL MPI_RECV( ibuf, 4, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
587	ierr )
588	hor_index_bounds(:,i) = ibuf(1:4)
589	ENDDO
590
591	ELSE
592	!
593	!-- Send index bounds to PE0
594	ibuf(1) = nxl
595	ibuf(2) = nxr
596	ibuf(3) = nys
597	ibuf(4) = nyn
598	CALL MPI_SEND( ibuf, 4, MPI_INTEGER, 0, myid, comm2d, ierr )
599
600	ENDIF
601
602
603	#if defined( __print )
604	!
605	!-- Control output
606	IF ( myid == 0 ) THEN
607	PRINT, ' processor topology *'
608	PRINT*, ' '
609	PRINT*, 'myid pcoord left right south north idx idy nxl: nxr',&
610	&' nys: nyn'
611	PRINT*, '------------------------------------------------------------',&
612	&'-----------'
613	WRITE (*,1000) 0, pcoord(1), pcoord(2), pleft, pright, psouth, pnorth, &
614	myidx, myidy, nxl, nxr, nys, nyn
615	1000 FORMAT (I4,2X,'(',I3,',',I3,')',3X,I4,2X,I4,3X,I4,2X,I4,2X,I3,1X,I3, &
616	2(2X,I4,':',I4))
617
618	!
619	!-- Receive data from the other PEs
620	DO i = 1,numprocs-1
621	CALL MPI_RECV( ibuf, 12, MPI_INTEGER, i, MPI_ANY_TAG, comm2d, status, &
622	ierr )
623	WRITE (*,1000) i, ( ibuf(j) , j = 1,12 )
624	ENDDO
625	ELSE
626
627	!
628	!-- Send data to PE0
629	ibuf(1) = pcoord(1); ibuf(2) = pcoord(2); ibuf(3) = pleft
630	ibuf(4) = pright; ibuf(5) = psouth; ibuf(6) = pnorth; ibuf(7) = myidx
631	ibuf(8) = myidy; ibuf(9) = nxl; ibuf(10) = nxr; ibuf(11) = nys
632	ibuf(12) = nyn
633	CALL MPI_SEND( ibuf, 12, MPI_INTEGER, 0, myid, comm2d, ierr )
634	ENDIF
635	#endif
636
637	!
638	!-- Determine the number of ghost point layers
639	IF ( scalar_advec == 'ws-scheme' .OR. &
640	momentum_advec == 'ws-scheme' .OR. nested_run ) THEN
641	nbgp = 3
642	ELSE
643	nbgp = 1
644	ENDIF
645
646	!
647	!-- Check that the number of computational grid points is not smaller than the number of
648	!-- ghost points.
649	IF ( nnx < nbgp ) THEN
650	WRITE( message_string, * ) 'number of subdomain grid points along x (', nnx, ') is smaller',&
651	'than the number of ghost points (', nbgp, ')'
652	CALL message( 'init_pegrid', 'PA0682', 1, 2, 0, 6, 0 )
653	ENDIF
654	IF ( nny < nbgp ) THEN
655	WRITE( message_string, * ) 'number of subdomain grid points along y (', nny, ') is smaller',&
656	'than the number of ghost points (', nbgp, ')'
657	CALL message( 'init_pegrid', 'PA0683', 1, 2, 0, 6, 0 )
658	ENDIF
659
660	!
661	!-- Create a new MPI derived datatype for the exchange of surface (xy) data,
662	!-- which is needed for coupled atmosphere-ocean runs.
663	!-- First, calculate number of grid points of an xy-plane.
664	ngp_xy = ( nxr - nxl + 1 + 2 * nbgp ) * ( nyn - nys + 1 + 2 * nbgp )
665	CALL MPI_TYPE_VECTOR( ngp_xy, 1, nzt-nzb+2, MPI_REAL, type_xy, ierr )
666	CALL MPI_TYPE_COMMIT( type_xy, ierr )
667
668	IF ( TRIM( coupling_mode ) /= 'uncoupled' .AND. .NOT. vnested ) THEN
669
670	!
671	!-- Pass the number of grid points of the atmosphere model to
672	!-- the ocean model and vice versa
673	IF ( coupling_mode == 'atmosphere_to_ocean' ) THEN
674
675	nx_a = nx
676	ny_a = ny
677
678	IF ( myid == 0 ) THEN
679
680	CALL MPI_SEND( nx_a, 1, MPI_INTEGER, numprocs, 1, comm_inter, &
681	ierr )
682	CALL MPI_SEND( ny_a, 1, MPI_INTEGER, numprocs, 2, comm_inter, &
683	ierr )
684	CALL MPI_SEND( pdims, 2, MPI_INTEGER, numprocs, 3, comm_inter, &
685	ierr )
686	CALL MPI_RECV( nx_o, 1, MPI_INTEGER, numprocs, 4, comm_inter, &
687	status, ierr )
688	CALL MPI_RECV( ny_o, 1, MPI_INTEGER, numprocs, 5, comm_inter, &
689	status, ierr )
690	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, numprocs, 6, &
691	comm_inter, status, ierr )
692	ENDIF
693
694	CALL MPI_BCAST( nx_o, 1, MPI_INTEGER, 0, comm2d, ierr )
695	CALL MPI_BCAST( ny_o, 1, MPI_INTEGER, 0, comm2d, ierr )
696	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr )
697
698	ELSEIF ( coupling_mode == 'ocean_to_atmosphere' ) THEN
699
700	nx_o = nx
701	ny_o = ny
702
703	IF ( myid == 0 ) THEN
704
705	CALL MPI_RECV( nx_a, 1, MPI_INTEGER, 0, 1, comm_inter, status, &
706	ierr )
707	CALL MPI_RECV( ny_a, 1, MPI_INTEGER, 0, 2, comm_inter, status, &
708	ierr )
709	CALL MPI_RECV( pdims_remote, 2, MPI_INTEGER, 0, 3, comm_inter, &
710	status, ierr )
711	CALL MPI_SEND( nx_o, 1, MPI_INTEGER, 0, 4, comm_inter, ierr )
712	CALL MPI_SEND( ny_o, 1, MPI_INTEGER, 0, 5, comm_inter, ierr )
713	CALL MPI_SEND( pdims, 2, MPI_INTEGER, 0, 6, comm_inter, ierr )
714	ENDIF
715
716	CALL MPI_BCAST( nx_a, 1, MPI_INTEGER, 0, comm2d, ierr)
717	CALL MPI_BCAST( ny_a, 1, MPI_INTEGER, 0, comm2d, ierr)
718	CALL MPI_BCAST( pdims_remote, 2, MPI_INTEGER, 0, comm2d, ierr)
719
720	ENDIF
721
722	ngp_a = ( nx_a+1 + 2 * nbgp ) * ( ny_a+1 + 2 * nbgp )
723	ngp_o = ( nx_o+1 + 2 * nbgp ) * ( ny_o+1 + 2 * nbgp )
724
725	!
726	!-- Determine if the horizontal grid and the number of PEs in ocean and
727	!-- atmosphere is same or not
728	IF ( nx_o == nx_a .AND. ny_o == ny_a .AND. &
729	pdims(1) == pdims_remote(1) .AND. pdims(2) == pdims_remote(2) ) &
730	THEN
731	coupling_topology = 0
732	ELSE
733	coupling_topology = 1
734	ENDIF
735
736	!
737	!-- Determine the target PEs for the exchange between ocean and
738	!-- atmosphere (comm2d)
739	IF ( coupling_topology == 0 ) THEN
740	!
741	!-- In case of identical topologies, every atmosphere PE has exactly one
742	!-- ocean PE counterpart and vice versa
743	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
744	target_id = myid + numprocs
745	ELSE
746	target_id = myid
747	ENDIF
748
749	ELSE
750	!
751	!-- In case of nonequivalent topology in ocean and atmosphere only for
752	!-- PE0 in ocean and PE0 in atmosphere a target_id is needed, since
753	!-- data echxchange between ocean and atmosphere will be done only
754	!-- between these PEs.
755	IF ( myid == 0 ) THEN
756
757	IF ( TRIM( coupling_mode ) == 'atmosphere_to_ocean' ) THEN
758	target_id = numprocs
759	ELSE
760	target_id = 0
761	ENDIF
762
763	ENDIF
764
765	ENDIF
766
767	ENDIF
768
769	!
770	!-- Store partner grid point co-ordinates as lists.
771	!-- Create custom MPI vector datatypes for contiguous data transfer
772	IF ( vnested ) CALL vnest_init_pegrid_domain
773
774	#else
775
776	!
777	!-- Array bounds when running on a single PE (respectively a non-parallel
778	!-- machine)
779	nxl = 0
780	nxr = nx
781	nnx = nxr - nxl + 1
782	nys = 0
783	nyn = ny
784	nny = nyn - nys + 1
785	nzb = 0
786	nzt = nz
787	nnz = nz
788
789	ALLOCATE( hor_index_bounds(4,0:0) )
790	hor_index_bounds(1,0) = nxl
791	hor_index_bounds(2,0) = nxr
792	hor_index_bounds(3,0) = nys
793	hor_index_bounds(4,0) = nyn
794
795	!
796	!-- Array bounds for the pressure solver (in the parallel code, these bounds
797	!-- are the ones for the transposed arrays)
798	nys_x = nys
799	nyn_x = nyn
800	nzb_x = nzb + 1
801	nzt_x = nzt
802
803	nxl_y = nxl
804	nxr_y = nxr
805	nzb_y = nzb + 1
806	nzt_y = nzt
807
808	nxl_z = nxl
809	nxr_z = nxr
810	nys_z = nys
811	nyn_z = nyn
812
813	#endif
814
815	!
816	!-- Calculate number of grid levels necessary for the multigrid poisson solver
817	!-- as well as the gridpoint indices on each level
818	IF ( psolver(1:9) == 'multigrid' ) THEN
819
820	!
821	!-- First calculate number of possible grid levels for the subdomains
822	mg_levels_x = 1
823	mg_levels_y = 1
824	mg_levels_z = 1
825
826	i = nnx
827	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
828	i = i / 2
829	mg_levels_x = mg_levels_x + 1
830	ENDDO
831
832	j = nny
833	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
834	j = j / 2
835	mg_levels_y = mg_levels_y + 1
836	ENDDO
837
838	k = nz ! do not use nnz because it might be > nz due to transposition
839	! requirements
840	DO WHILE ( MOD( k, 2 ) == 0 .AND. k /= 2 )
841	k = k / 2
842	mg_levels_z = mg_levels_z + 1
843	ENDDO
844	!
845	!-- The optimized MG-solver does not allow odd values for nz at the coarsest
846	!-- grid level
847	IF ( TRIM( psolver ) /= 'multigrid_noopt' ) THEN
848	IF ( MOD( k, 2 ) /= 0 ) mg_levels_z = mg_levels_z - 1
849	!
850	!-- An odd value of nz does not work. The finest level must have an even
851	!-- value.
852	IF ( mg_levels_z == 0 ) THEN
853	message_string = 'optimized multigrid method requires nz to be even'
854	CALL message( 'init_pegrid', 'PA0495', 1, 2, 0, 6, 0 )
855	ENDIF
856	ENDIF
857
858	maximum_grid_level = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
859	!
860	!-- Check if subdomain sizes prevents any coarsening.
861	!-- This case, the maximum number of grid levels is 1, i.e. effectively
862	!-- a Gauss-Seidel scheme is applied rather than a multigrid approach.
863	!-- Give a warning in this case.
864	IF ( maximum_grid_level == 1 .AND. mg_switch_to_pe0_level == -1 ) THEN
865	message_string = 'No grid coarsening possible, multigrid ' // &
866	'approach effectively reduces to a Gauss-Seidel ' //&
867	'scheme.'
868
869	CALL message( 'poismg', 'PA0648', 0, 1, 0, 6, 0 )
870	ENDIF
871
872	!
873	!-- Find out, if the total domain allows more levels. These additional
874	!-- levels are identically processed on all PEs.
875	IF ( numprocs > 1 .AND. mg_switch_to_pe0_level /= -1 ) THEN
876
877	IF ( mg_levels_z > MIN( mg_levels_x, mg_levels_y ) ) THEN
878
879	mg_switch_to_pe0_level_l = maximum_grid_level
880
881	mg_levels_x = 1
882	mg_levels_y = 1
883
884	i = nx+1
885	DO WHILE ( MOD( i, 2 ) == 0 .AND. i /= 2 )
886	i = i / 2
887	mg_levels_x = mg_levels_x + 1
888	ENDDO
889
890	j = ny+1
891	DO WHILE ( MOD( j, 2 ) == 0 .AND. j /= 2 )
892	j = j / 2
893	mg_levels_y = mg_levels_y + 1
894	ENDDO
895
896	maximum_grid_level_l = MIN( mg_levels_x, mg_levels_y, mg_levels_z )
897
898	IF ( maximum_grid_level_l > mg_switch_to_pe0_level_l ) THEN
899	mg_switch_to_pe0_level_l = maximum_grid_level_l - &
900	mg_switch_to_pe0_level_l + 1
901	ELSE
902	mg_switch_to_pe0_level_l = 0
903	ENDIF
904
905	ELSE
906
907	mg_switch_to_pe0_level_l = 0
908	maximum_grid_level_l = maximum_grid_level
909
910	ENDIF
911
912	!
913	!-- Use switch level calculated above only if it is not pre-defined
914	!-- by user
915	IF ( mg_switch_to_pe0_level == 0 ) THEN
916	IF ( mg_switch_to_pe0_level_l /= 0 ) THEN
917	mg_switch_to_pe0_level = mg_switch_to_pe0_level_l
918	maximum_grid_level = maximum_grid_level_l
919	ENDIF
920
921	ELSE
922	!
923	!-- Check pre-defined value and reset to default, if neccessary
924	IF ( mg_switch_to_pe0_level < mg_switch_to_pe0_level_l .OR. &
925	mg_switch_to_pe0_level >= maximum_grid_level_l ) THEN
926	message_string = 'mg_switch_to_pe0_level ' // &
927	'out of range and reset to 0'
928	CALL message( 'init_pegrid', 'PA0235', 0, 1, 0, 6, 0 )
929	mg_switch_to_pe0_level = 0
930	ELSE
931	!
932	!-- Use the largest number of possible levels anyway and recalculate
933	!-- the switch level to this largest number of possible values
934	maximum_grid_level = maximum_grid_level_l
935
936	ENDIF
937
938	ENDIF
939
940	ENDIF
941
942	ALLOCATE( grid_level_count(maximum_grid_level), &
943	nxl_mg(0:maximum_grid_level), nxr_mg(0:maximum_grid_level), &
944	nyn_mg(0:maximum_grid_level), nys_mg(0:maximum_grid_level), &
945	nzt_mg(0:maximum_grid_level) )
946
947	grid_level_count = 0
948	!
949	!-- Index zero required as dummy due to definition of arrays f2 and p2 in
950	!-- recursive subroutine next_mg_level
951	nxl_mg(0) = 0; nxr_mg(0) = 0; nyn_mg(0) = 0; nys_mg(0) = 0; nzt_mg(0) = 0
952
953	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzt_l = nzt
954
955	DO i = maximum_grid_level, 1 , -1
956
957	IF ( i == mg_switch_to_pe0_level ) THEN
958	#if defined( __parallel )
959	!
960	!-- Save the grid size of the subdomain at the switch level, because
961	!-- it is needed in poismg.
962	ind(1) = nxl_l; ind(2) = nxr_l
963	ind(3) = nys_l; ind(4) = nyn_l
964	ind(5) = nzt_l
965	ALLOCATE( ind_all(5*numprocs), mg_loc_ind(5,0:numprocs-1) )
966	CALL MPI_ALLGATHER( ind, 5, MPI_INTEGER, ind_all, 5, &
967	MPI_INTEGER, comm2d, ierr )
968	DO j = 0, numprocs-1
969	DO k = 1, 5
970	mg_loc_ind(k,j) = ind_all(k+j*5)
971	ENDDO
972	ENDDO
973	DEALLOCATE( ind_all )
974	!
975	!-- Calculate the grid size of the total domain
976	nxr_l = ( nxr_l-nxl_l+1 ) * pdims(1) - 1
977	nxl_l = 0
978	nyn_l = ( nyn_l-nys_l+1 ) * pdims(2) - 1
979	nys_l = 0
980	!
981	!-- The size of this gathered array must not be larger than the
982	!-- array tend, which is used in the multigrid scheme as a temporary
983	!-- array. Therefore the subdomain size of an PE is calculated and
984	!-- the size of the gathered grid. These values are used in
985	!-- routines pres and poismg
986	subdomain_size = ( nxr - nxl + 2 * nbgp + 1 ) * &
987	( nyn - nys + 2 * nbgp + 1 ) * ( nzt - nzb + 2 )
988	gathered_size = ( nxr_l - nxl_l + 3 ) * ( nyn_l - nys_l + 3 ) * &
989	( nzt_l - nzb + 2 )
990
991	#else
992	message_string = 'multigrid gather/scatter impossible ' // &
993	'in non parallel mode'
994	CALL message( 'init_pegrid', 'PA0237', 1, 2, 0, 6, 0 )
995	#endif
996	ENDIF
997
998	nxl_mg(i) = nxl_l
999	nxr_mg(i) = nxr_l
1000	nys_mg(i) = nys_l
1001	nyn_mg(i) = nyn_l
1002	nzt_mg(i) = nzt_l
1003
1004	nxl_l = nxl_l / 2
1005	nxr_l = nxr_l / 2
1006	nys_l = nys_l / 2
1007	nyn_l = nyn_l / 2
1008	nzt_l = nzt_l / 2
1009
1010	ENDDO
1011
1012	!
1013	!-- Temporary problem: Currently calculation of maxerror in routine poismg crashes
1014	!-- if grid data are collected on PE0 already on the finest grid level.
1015	!-- To be solved later.
1016	IF ( maximum_grid_level == mg_switch_to_pe0_level ) THEN
1017	message_string = 'grid coarsening on subdomain level cannot be performed'
1018	CALL message( 'poismg', 'PA0236', 1, 2, 0, 6, 0 )
1019	ENDIF
1020
1021	ELSE
1022
1023	maximum_grid_level = 0
1024
1025	ENDIF
1026
1027	!
1028	!-- Default level 0 tells exchange_horiz that all ghost planes have to be
1029	!-- exchanged. grid_level is adjusted in poismg, where only one ghost plane
1030	!-- is required.
1031	grid_level = 0
1032
1033	#if defined( __parallel )
1034	!
1035	!-- Gridpoint number for the exchange of ghost points (y-line for 2D-arrays)
1036	ngp_y = nyn - nys + 1 + 2 * nbgp
1037
1038	!
1039	!-- Define new MPI derived datatypes for the exchange of ghost points in
1040	!-- x- and y-direction for 2D-arrays (line)
1041	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_REAL, type_x, &
1042	ierr )
1043	CALL MPI_TYPE_COMMIT( type_x, ierr )
1044
1045	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_REAL, type_y, ierr )
1046	CALL MPI_TYPE_COMMIT( type_y, ierr )
1047	!
1048	!-- Define new MPI derived datatypes for the exchange of ghost points in
1049	!-- x- and y-direction for 2D-INTEGER arrays (line) - on normal grid.
1050	!-- Define types for 32-bit and 8-bit Integer. The 8-bit Integer are only
1051	!-- required on normal grid, while 32-bit Integer may be also required on
1052	!-- coarser grid level in case of multigrid solver.
1053	!
1054	!-- 8-bit Integer
1055	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_BYTE, &
1056	type_x_byte, ierr )
1057	CALL MPI_TYPE_COMMIT( type_x_byte, ierr )
1058
1059	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_BYTE, &
1060	type_y_byte, ierr )
1061	CALL MPI_TYPE_COMMIT( type_y_byte, ierr )
1062	!
1063	!-- 32-bit Integer
1064	ALLOCATE( type_x_int(0:maximum_grid_level), &
1065	type_y_int(0:maximum_grid_level) )
1066
1067	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2*nbgp, nbgp, ngp_y, MPI_INTEGER, &
1068	type_x_int(0), ierr )
1069	CALL MPI_TYPE_COMMIT( type_x_int(0), ierr )
1070
1071	CALL MPI_TYPE_VECTOR( nbgp, ngp_y, ngp_y, MPI_INTEGER, type_y_int(0), ierr )
1072	CALL MPI_TYPE_COMMIT( type_y_int(0), ierr )
1073	!
1074	!-- Calculate gridpoint numbers for the exchange of ghost points along x
1075	!-- (yz-plane for 3D-arrays) and define MPI derived data type(s) for the
1076	!-- exchange of ghost points in y-direction (xz-plane).
1077	!-- Do these calculations for the model grid and (if necessary) also
1078	!-- for the coarser grid levels used in the multigrid method
1079	ALLOCATE ( ngp_xz(0:maximum_grid_level), &
1080	ngp_xz_int(0:maximum_grid_level), &
1081	ngp_yz(0:maximum_grid_level), &
1082	ngp_yz_int(0:maximum_grid_level), &
1083	type_xz(0:maximum_grid_level), &
1084	type_xz_int(0:maximum_grid_level), &
1085	type_yz(0:maximum_grid_level), &
1086	type_yz_int(0:maximum_grid_level) )
1087
1088	nxl_l = nxl; nxr_l = nxr; nys_l = nys; nyn_l = nyn; nzb_l = nzb; nzt_l = nzt
1089
1090	!
1091	!-- Discern between the model grid, which needs nbgp ghost points and
1092	!-- grid levels for the multigrid scheme. In the latter case only one
1093	!-- ghost point is necessary.
1094	!-- First definition of MPI-datatypes for exchange of ghost layers on normal
1095	!-- grid. The following loop is needed for data exchange in poismg.f90.
1096	!
1097	!-- Determine number of grid points of yz-layer for exchange
1098	ngp_yz(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1099
1100	!
1101	!-- Define an MPI-datatype for the exchange of left/right boundaries.
1102	!-- Although data are contiguous in physical memory (which does not
1103	!-- necessarily require an MPI-derived datatype), the data exchange between
1104	!-- left and right PE's using the MPI-derived type is 10% faster than without.
1105	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz(0), &
1106	MPI_REAL, type_xz(0), ierr )
1107	CALL MPI_TYPE_COMMIT( type_xz(0), ierr )
1108
1109	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz(0), ngp_yz(0), MPI_REAL, type_yz(0), &
1110	ierr )
1111	CALL MPI_TYPE_COMMIT( type_yz(0), ierr )
1112
1113	!
1114	!-- Define data types for exchange of 3D Integer arrays.
1115	ngp_yz_int(0) = (nzt - nzb + 2) * (nyn - nys + 1 + 2 * nbgp)
1116
1117	CALL MPI_TYPE_VECTOR( nxr-nxl+1+2nbgp, nbgp(nzt-nzb+2), ngp_yz_int(0), &
1118	MPI_INTEGER, type_xz_int(0), ierr )
1119	CALL MPI_TYPE_COMMIT( type_xz_int(0), ierr )
1120
1121	CALL MPI_TYPE_VECTOR( nbgp, ngp_yz_int(0), ngp_yz_int(0), MPI_INTEGER, &
1122	type_yz_int(0), ierr )
1123	CALL MPI_TYPE_COMMIT( type_yz_int(0), ierr )
1124
1125	!
1126	!-- Definition of MPI-datatypes for multigrid method (coarser level grids)
1127	IF ( psolver(1:9) == 'multigrid' ) THEN
1128	!
1129	!-- Definition of MPI-datatyoe as above, but only 1 ghost level is used
1130	DO i = maximum_grid_level, 1 , -1
1131	!
1132	!-- For 3D-exchange on different multigrid level, one ghost point for
1133	!-- REAL arrays, two ghost points for INTEGER arrays
1134	ngp_xz(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
1135	ngp_yz(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
1136
1137	ngp_xz_int(i) = (nzt_l - nzb_l + 2) * (nxr_l - nxl_l + 3)
1138	ngp_yz_int(i) = (nzt_l - nzb_l + 2) * (nyn_l - nys_l + 3)
1139	!
1140	!-- MPI data type for REAL arrays, for xz-layers
1141	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz(i), &
1142	MPI_REAL, type_xz(i), ierr )
1143	CALL MPI_TYPE_COMMIT( type_xz(i), ierr )
1144
1145	!
1146	!-- MPI data type for INTEGER arrays, for xz-layers
1147	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+3, nzt_l-nzb_l+2, ngp_yz_int(i), &
1148	MPI_INTEGER, type_xz_int(i), ierr )
1149	CALL MPI_TYPE_COMMIT( type_xz_int(i), ierr )
1150
1151	!
1152	!-- MPI data type for REAL arrays, for yz-layers
1153	CALL MPI_TYPE_VECTOR( 1, ngp_yz(i), ngp_yz(i), MPI_REAL, type_yz(i), &
1154	ierr )
1155	CALL MPI_TYPE_COMMIT( type_yz(i), ierr )
1156	!
1157	!-- MPI data type for INTEGER arrays, for yz-layers
1158	CALL MPI_TYPE_VECTOR( 1, ngp_yz_int(i), ngp_yz_int(i), MPI_INTEGER, &
1159	type_yz_int(i), ierr )
1160	CALL MPI_TYPE_COMMIT( type_yz_int(i), ierr )
1161
1162
1163	!-- For 2D-exchange of INTEGER arrays on coarser grid level, where 2 ghost
1164	!-- points need to be exchanged. Only required for 32-bit Integer arrays.
1165	CALL MPI_TYPE_VECTOR( nxr_l-nxl_l+5, 2, nyn_l-nys_l+5, MPI_INTEGER, &
1166	type_x_int(i), ierr )
1167	CALL MPI_TYPE_COMMIT( type_x_int(i), ierr )
1168
1169
1170	CALL MPI_TYPE_VECTOR( 2, nyn_l-nys_l+5, nyn_l-nys_l+5, MPI_INTEGER, &
1171	type_y_int(i), ierr )
1172	CALL MPI_TYPE_COMMIT( type_y_int(i), ierr )
1173
1174	nxl_l = nxl_l / 2
1175	nxr_l = nxr_l / 2
1176	nys_l = nys_l / 2
1177	nyn_l = nyn_l / 2
1178	nzt_l = nzt_l / 2
1179
1180	ENDDO
1181
1182	ENDIF
1183
1184	#endif
1185
1186	#if defined( __parallel )
1187	!
1188	!-- Setting of flags for inflow/outflow/nesting conditions.
1189	IF ( pleft == MPI_PROC_NULL ) THEN
1190	IF ( bc_lr == 'dirichlet/radiation' .OR. bc_lr == 'nested' .OR. &
1191	bc_lr == 'nesting_offline' ) THEN
1192	bc_dirichlet_l = .TRUE.
1193	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1194	bc_radiation_l = .TRUE.
1195	ENDIF
1196	ENDIF
1197
1198	IF ( pright == MPI_PROC_NULL ) THEN
1199	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1200	bc_radiation_r = .TRUE.
1201	ELSEIF ( bc_lr == 'radiation/dirichlet' .OR. bc_lr == 'nested' .OR. &
1202	bc_lr == 'nesting_offline' ) THEN
1203	bc_dirichlet_r = .TRUE.
1204	ENDIF
1205	ENDIF
1206
1207	IF ( psouth == MPI_PROC_NULL ) THEN
1208	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1209	bc_radiation_s = .TRUE.
1210	ELSEIF ( bc_ns == 'radiation/dirichlet' .OR. bc_ns == 'nested' .OR. &
1211	bc_ns == 'nesting_offline' ) THEN
1212	bc_dirichlet_s = .TRUE.
1213	ENDIF
1214	ENDIF
1215
1216	IF ( pnorth == MPI_PROC_NULL ) THEN
1217	IF ( bc_ns == 'dirichlet/radiation' .OR. bc_ns == 'nested' .OR. &
1218	bc_ns == 'nesting_offline' ) THEN
1219	bc_dirichlet_n = .TRUE.
1220	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1221	bc_radiation_n = .TRUE.
1222	ENDIF
1223	ENDIF
1224	!
1225	!-- In case of synthetic turbulence geneartor determine ids.
1226	!-- Please note, if no forcing or nesting is applied, the generator is applied
1227	!-- only at the left lateral boundary.
1228	IF ( use_syn_turb_gen ) THEN
1229	IF ( bc_dirichlet_l ) THEN
1230	id_stg_left_l = myidx
1231	ELSE
1232	id_stg_left_l = 0
1233	ENDIF
1234	IF ( bc_dirichlet_r ) THEN
1235	id_stg_right_l = myidx
1236	ELSE
1237	id_stg_right_l = 0
1238	ENDIF
1239	IF ( bc_dirichlet_s ) THEN
1240	id_stg_south_l = myidy
1241	ELSE
1242	id_stg_south_l = 0
1243	ENDIF
1244	IF ( bc_dirichlet_n ) THEN
1245	id_stg_north_l = myidy
1246	ELSE
1247	id_stg_north_l = 0
1248	ENDIF
1249
1250	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1251	CALL MPI_ALLREDUCE( id_stg_left_l, id_stg_left, 1, MPI_INTEGER, &
1252	MPI_SUM, comm1dx, ierr )
1253
1254	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1255	CALL MPI_ALLREDUCE( id_stg_right_l, id_stg_right, 1, MPI_INTEGER, &
1256	MPI_SUM, comm1dx, ierr )
1257
1258	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1259	CALL MPI_ALLREDUCE( id_stg_south_l, id_stg_south, 1, MPI_INTEGER, &
1260	MPI_SUM, comm1dy, ierr )
1261
1262	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1263	CALL MPI_ALLREDUCE( id_stg_north_l, id_stg_north, 1, MPI_INTEGER, &
1264	MPI_SUM, comm1dy, ierr )
1265
1266	ENDIF
1267
1268	!
1269	!-- Broadcast the id of the inflow PE
1270	IF ( bc_dirichlet_l ) THEN
1271	id_inflow_l = myidx
1272	ELSE
1273	id_inflow_l = 0
1274	ENDIF
1275	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1276	CALL MPI_ALLREDUCE( id_inflow_l, id_inflow, 1, MPI_INTEGER, MPI_SUM, &
1277	comm1dx, ierr )
1278
1279	!
1280	!-- Broadcast the id of the recycling plane
1281	!-- WARNING: needs to be adjusted in case of inflows other than from left side!
1282	IF ( turbulent_inflow ) THEN
1283
1284	IF ( NINT( recycling_width / dx, KIND=idp ) >= nxl .AND. &
1285	NINT( recycling_width / dx, KIND=idp ) <= nxr ) THEN
1286	id_recycling_l = myidx
1287	ELSE
1288	id_recycling_l = 0
1289	ENDIF
1290	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1291	CALL MPI_ALLREDUCE( id_recycling_l, id_recycling, 1, MPI_INTEGER, MPI_SUM, &
1292	comm1dx, ierr )
1293
1294	ENDIF
1295
1296	!
1297	!-- Broadcast the id of the outflow PE and outflow-source plane
1298	IF ( turbulent_outflow ) THEN
1299
1300	IF ( bc_radiation_r ) THEN
1301	id_outflow_l = myidx
1302	ELSE
1303	id_outflow_l = 0
1304	ENDIF
1305	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1306	CALL MPI_ALLREDUCE( id_outflow_l, id_outflow, 1, MPI_INTEGER, MPI_SUM, &
1307	comm1dx, ierr )
1308
1309	IF ( NINT( outflow_source_plane / dx ) >= nxl .AND. &
1310	NINT( outflow_source_plane / dx ) <= nxr ) THEN
1311	id_outflow_source_l = myidx
1312	ELSE
1313	id_outflow_source_l = 0
1314	ENDIF
1315	IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr )
1316	CALL MPI_ALLREDUCE( id_outflow_source_l, id_outflow_source, 1, &
1317	MPI_INTEGER, MPI_SUM, comm1dx, ierr )
1318
1319	ENDIF
1320
1321	CALL location_message( 'creating virtual PE grids + MPI derived data types', 'finished' )
1322
1323	#else
1324	IF ( bc_lr == 'dirichlet/radiation' ) THEN
1325	bc_dirichlet_l = .TRUE.
1326	bc_radiation_r = .TRUE.
1327	ELSEIF ( bc_lr == 'radiation/dirichlet' ) THEN
1328	bc_radiation_l = .TRUE.
1329	bc_dirichlet_r = .TRUE.
1330	ENDIF
1331
1332	IF ( bc_ns == 'dirichlet/radiation' ) THEN
1333	bc_dirichlet_n = .TRUE.
1334	bc_radiation_s = .TRUE.
1335	ELSEIF ( bc_ns == 'radiation/dirichlet' ) THEN
1336	bc_radiation_n = .TRUE.
1337	bc_dirichlet_s = .TRUE.
1338	ENDIF
1339	#endif
1340
1341	!
1342	!-- At the inflow or outflow, u or v, respectively, have to be calculated for
1343	!-- one more grid point.
1344	IF ( bc_dirichlet_l .OR. bc_radiation_l ) THEN
1345	nxlu = nxl + 1
1346	ELSE
1347	nxlu = nxl
1348	ENDIF
1349	IF ( bc_dirichlet_s .OR. bc_radiation_s ) THEN
1350	nysv = nys + 1
1351	ELSE
1352	nysv = nys
1353	ENDIF
1354
1355	!
1356	!-- Allocate wall flag arrays used in the multigrid solver
1357	IF ( psolver(1:9) == 'multigrid' ) THEN
1358
1359	DO i = maximum_grid_level, 1, -1
1360
1361	SELECT CASE ( i )
1362
1363	CASE ( 1 )
1364	ALLOCATE( wall_flags_1(nzb:nzt_mg(i)+1, &
1365	nys_mg(i)-1:nyn_mg(i)+1, &
1366	nxl_mg(i)-1:nxr_mg(i)+1) )
1367
1368	CASE ( 2 )
1369	ALLOCATE( wall_flags_2(nzb:nzt_mg(i)+1, &
1370	nys_mg(i)-1:nyn_mg(i)+1, &
1371	nxl_mg(i)-1:nxr_mg(i)+1) )
1372
1373	CASE ( 3 )
1374	ALLOCATE( wall_flags_3(nzb:nzt_mg(i)+1, &
1375	nys_mg(i)-1:nyn_mg(i)+1, &
1376	nxl_mg(i)-1:nxr_mg(i)+1) )
1377
1378	CASE ( 4 )
1379	ALLOCATE( wall_flags_4(nzb:nzt_mg(i)+1, &
1380	nys_mg(i)-1:nyn_mg(i)+1, &
1381	nxl_mg(i)-1:nxr_mg(i)+1) )
1382
1383	CASE ( 5 )
1384	ALLOCATE( wall_flags_5(nzb:nzt_mg(i)+1, &
1385	nys_mg(i)-1:nyn_mg(i)+1, &
1386	nxl_mg(i)-1:nxr_mg(i)+1) )
1387
1388	CASE ( 6 )
1389	ALLOCATE( wall_flags_6(nzb:nzt_mg(i)+1, &
1390	nys_mg(i)-1:nyn_mg(i)+1, &
1391	nxl_mg(i)-1:nxr_mg(i)+1) )
1392
1393	CASE ( 7 )
1394	ALLOCATE( wall_flags_7(nzb:nzt_mg(i)+1, &
1395	nys_mg(i)-1:nyn_mg(i)+1, &
1396	nxl_mg(i)-1:nxr_mg(i)+1) )
1397
1398	CASE ( 8 )
1399	ALLOCATE( wall_flags_8(nzb:nzt_mg(i)+1, &
1400	nys_mg(i)-1:nyn_mg(i)+1, &
1401	nxl_mg(i)-1:nxr_mg(i)+1) )
1402
1403	CASE ( 9 )
1404	ALLOCATE( wall_flags_9(nzb:nzt_mg(i)+1, &
1405	nys_mg(i)-1:nyn_mg(i)+1, &
1406	nxl_mg(i)-1:nxr_mg(i)+1) )
1407
1408	CASE ( 10 )
1409	ALLOCATE( wall_flags_10(nzb:nzt_mg(i)+1, &
1410	nys_mg(i)-1:nyn_mg(i)+1, &
1411	nxl_mg(i)-1:nxr_mg(i)+1) )
1412
1413	CASE DEFAULT
1414	message_string = 'more than 10 multigrid levels'
1415	CALL message( 'init_pegrid', 'PA0238', 1, 2, 0, 6, 0 )
1416
1417	END SELECT
1418
1419	ENDDO
1420
1421	ENDIF
1422
1423	END SUBROUTINE init_pegrid

Note: See TracBrowser for help on using the repository browser.

Download in other formats:

| Impressum | ©Leibniz Universität Hannover |